About the client
CardiaTec is an AI-driven drug target discovery company specialising in cardiovascular disease. They’re applying artificial intelligence on multi-omics dataset of human cardiac tissue to discover the next generation of medicines for cardiovascular disease.
Business challenge
Complex data analysis and visualisation challenges in
multi-omics data
CardiaTec has an AI team that is building a proprietary multi-omics dataset of human cardiac tissue.
This AI platform integrates diverse biological data layers, enabling a comprehensive understanding of cardiac disease biology.
However, this complex dataset was not being translated into actionable knowledge because of the lack of data analysis and visualisation around it.
CardiaTec’s researchers were unable to extract meaningful insights from the data, crucial for advancing cardiac health research and developing new treatments.
The complexity of the data, coupled with its volume, resulted in the following issues:
- Difficulty in visualising and analysing extensive multi-omic data to extract meaningful insights for drug target development.
- The visualisation of data on the graph rendered the data nearly impossible to decipher, obscuring critical insights and correlations.
CardiaTec reached out to GoodCore to build a web-based data analysis and visualisation platform to assist its in-house researchers extract insights from the multi-omics data.
The solution
Project goals
The project entailed the following requirements:
- Develop a web-based data visualisation and reporting platform to uncover relationships across every layer of biological data.
- Dynamically integrate the enormous data from their AI platform, ensuring accuracy and constant updates.
- Implement network visualisation to display the interactions between various biological entities in 3-dimensional space.
Product development challenges and solutions
Challenge #1: Complex network visualisation and interaction
The main challenge in this project was the 3D visualisation of a complex network of protein interactions. This network not only needed to be detailed and comprehensive but also required each node to be interactive, containing various feature sets. The requirement was to arrange nodes in a visually appealing and informative layout, with related nodes clustering each other, so that researchers could navigate the extensive interactions within the protein network.
Solution: 3D force-directed graph with customisation
To overcome this challenge, we used a graph data structure in a 3-dimensional space using a force-directed iterative layout to reveal complex relationships and interactions across various biological layers. Initially, Cryptoscape was considered for network visualisation due to its capabilities in charting. However, its separate app requirement for 3D rendering led us to choose the 3D force graph open-source tool, which we heavily customised to adhere to CardiaTec’s specific business rules. The 3D force graph provided all the essential functionalities for effective visualisation. Its ability to group the objects naturally connected to each other made it possible to discover subtle relationships between groups.
Challenge #2: Ensuring intuitive UX for complex data
The application’s UX had to effectively manage the vast amount of information across multiple diseases and datasets. With the intricate analysis required for each disease, it was crucial to design a UX that was not only comprehensive but also easy to navigate. So that researchers could efficiently access and interpret the data without being overwhelmed by its complexity.
Solution: Enhanced UX with filters and visualisation
To achieve this, we introduced multiple filters, enabling researchers to easily analyse disease-specific targets. The protein details dashboard was designed to present high-level protein information in a digestible format, utilising tables and graphs for clarity. For the visualisation of protein interactions, we employed different shapes for metabolic proteins and colour-coded nodes according to their expressions, enhancing the visual differentiation of data. Additional functionalities like 2D viewing, rotation options, and label toggling were integrated to increase the application's intuitiveness and ease of use and improve the overall efficiency of the research process.
Challenge #3: Data volume and processing speed
Managing the huge volume of data and ensuring speedy processing was another challenge. There were millions of records for each test and each file to be uploaded had intricate links to other files, necessitating a validation of all these interconnections before any upload could proceed. Additionally, the data processing aspect demanded high efficiency. It was imperative to optimise queries to ensure that the graphical representation of data could be generated instantaneously.
Solution: Python tech stack and database optimisation
To address the challenge, we used the Python tech stack because of its robust set of libraries and data parsing efficiency to swiftly upload millions of records in the system. On the visualisation front, we optimised the database. This involved de-normalising the data to enhance the speed and efficiency of queries. Careful indexing was also implemented to expedite search operations and make the queries as fast as possible.
Challenge #4: Startup budget constraints
CardiaTec was a startup in its pre-seed investment phase and working with a limited budget. The financial constraints required a customised approach to development, necessitating careful consideration of how to maximise value while minimising costs.
Solution: Feature prioritisation and flexible development approach
Our strategy during the discovery phase was to prioritise features and define the project scope to align with CardiaTec's budget. However, as the project progressed, the scope evolved – for instance, CardiaTec's initial requirement for analyses of a single disease expanded to include multiple diseases. At GoodCore, we have a philosophy of flexibility and adaptability, especially when it involves supporting our client's evolving needs. We managed the project within the available budget, squeezing in essential features and scope changes where possible. This flexible and client-focused approach enabled us to deliver a successful solution to CardiaTec, despite the financial constraints of a startup environment.
Their project management skills were excellent. GoodCore Software kept within the schedule and delivered everything on time, even though we asked for changes a handful of times. Moreover, the GoodCore Software team was on top of any issues, handling them within a week, which we really appreciated.
Woochang Hwang, Senior Machine Learning Engineer,
CardiaTec
The impact
The platform developed by GoodCore has transformed CardiaTec's research approach and capabilities, leading to advancements in the field of cardiovascular disease research.
Before
Limited understanding of complex protein networks; reliance on 2D representations.
Difficulty in interpreting relationships between data points due to inadequate visualisation tools.
Challenges in studying connectivity and centrality in molecular interactions.
Limited insights into behavioural patterns and interactions at individual and group levels.
After
Enhanced comprehension of complex networks with 3D visualisations revealing new patterns and clusters.
Advanced visualisation tools providing clear insights into relationships, leading to new drug and target discoveries.
Improved ability to analyse molecular networks, influencing the development of targeted therapies.
In-depth analysis of behavioural patterns, informing personalised treatment strategies.
Through this new understanding of disease biology and exploration of target-drug interactions, CardiaTec’s approach will deliver new treatments to patients faster, cheaper and with a reduced risk of failure.
Wondering about the price range for a similar solution?
Get ballpark estimates and insights into team structure and timelines.